Image registration aims at finding an optimal transformation between different representations of one or more objects, i.e. between different images. Registration techniques can be useful in medical procedures in which a pre-operative image space may need to be properly correlated to a real-time physical space. In image-guided surgical procedures, for example, pre-operatively acquired images may have to be registered onto intra-operative, near real-time images. In this way, the surgeon can be guided during his operation by viewing, in real time, images of the anatomical region being treated and/or the surgical devices. In practice, a formal mathematical transformation may be determined that best aligns the pre-operative image coordinate system with the patient's physical world coordinate system, defined for example in the treatment room.
The registration of preoperative 3D images onto real-time 2D projection images (e.g. 2D x-ray projection images) is often referred to as “2D-3D image registration.” 2D-3D image registration is widely used in image-guided surgical procedures. Because real-time x-ray images in general are merely 2D projections, the lack of 3D information can hinder accurate surgical guidance. Pre-operative 3D scans (e.g. CT scans or MRI scans) of the target region can provide the necessary 3D information.
A robust and accurate 2D-3D registration algorithm is needed in order for the position of the anatomical target (and/or relevant surgical instruments), as viewed on the real-time 2D images, to be reliably correlated to their position as visualized through the pre-operative 3D scans. As one example, during radiotherapy or radiosurgery, 2D-3D registration can be used to properly direct radiation onto a tumorous target that is visible in the images. As another example, in a surgical navigation system, 2D-3D registration can be used to track in real time the changing position of a surgical probe on a display of the preoperative images.
A known registration method is to identify corresponding features in each coordinate system. For example, fiducial markers may be attached to or implanted in the patient before the pre-operative images are acquired, for point-based alignment. The markers may be tracked using an optical localization device. Typically, these fiducial markers may be designed so that they can be accurately localized in the pre-operative image as well as in the physical world. The respective localization points may then used to calculate a rigid body transformation between the two coordinate systems.
Fiducials-based tracking can be difficult for the patient, for a number of reasons. For example, high accuracy tends to be achieved by using bone-implanted fiducial markers, but less invasive techniques such as skin-attached markers or anatomical positions tend to be less accurate. Implantation of fiducials into a patient may often be painful and difficult, especially for the C-spine, the implantation process for which may frequently lead to clinical complications.
A number of attempts have been made in the art to develop techniques for fiducial-less tracking. These known methods generally assume a rigid body transformation, i.e. a rigid body rotation and a rigid body translation. Such a rigid transformation typically ignores local variations during the transformation, and assumes that the patient's anatomy is a rigid body, and that all of the rigid body constraints should be preserved. A lot of clinical data has shown that the rigid transformation model may be inadequate in many cases. For example, although the deformation of an individual bone may be relatively accurately modeled as a rigid transformation, adjacent bones could move relatively to the individual bone, and soft tissue surrounding the bones could deform. Accordingly, non-rigid registration algorithms may be required in order to account for real patient body deformation, and thus track an anatomical region more precisely.
Another undesirable feature of these known fiducial-less tracking methods is the requirement of user interaction. Moreover, the accuracy resulting from these known methods is generally insufficient, the errors typically tending to be greater than about 1 mm. Further, the results can be unstable, and may be sensitive to an initial guess.
It is desirable that a method and system for performing 2D/3D medical image registration be provided which does not require the use of fiducial marks, does not require user interaction, does not assume a rigid transformation, and uses as little computing time as possible.